# Terrain, tide datum and historical-event validation

Independent research audit, 20 October 2026. No model inputs were changed. The available evidence supports using the 2023 terrain as a regional foundation; it does not establish current, continuous seawall crests or property-level flood accuracy.

## Fresh catalog check

A fresh [USGS TNM products API](https://tnmaccess.nationalmap.gov/api/v1/products?bbox=-117.94%2C33.58%2C-117.86%2C33.63&datasets=Lidar%20Point%20Cloud%20%28LPC%29&max=1000) query for the Peninsula/Islands envelope returned 89 lidar tiles. The newest named acquisition is CA_CaliforniaGaps_B23, whose tile publication dates are August 2025. Publication is not acquisition. The saved complete response is `terrain-catalog-tnm-20261020.json`. Within this catalog and footprint, no newer airborne lidar acquisition was identified. This is bounded negative evidence, not proof that no unpublicized survey exists.

A matching [NOAA interagency inventory query](https://www.coast.noaa.gov/arcgis/rest/services/USInteragencyElevationInventory/USInteragencyElevInventory/MapServer/1) returned 18 intersecting projects, saved in `terrain-catalog-noaa-20261020.json`. That inventory did not include the current 2023 Gaps project, so it cannot independently establish the newest coverage. NOAA's separate [InPort record](https://www.fisheries.noaa.gov/inport/item/77693) does document it.

The current source work unit CA_CaliforniaGaps_2_B23/300619 was acquired 24 October–11 November 2023. Its original DEM has 0.5 m spacing, EPSG 6340 horizontal coordinates, NAVD88 heights and GEOID18, and hydro-flattened water. These are acquisition-era surface observations, not a 2026 survey. [USGS project report, 9 May 2025](https://prd-tnm.s3.amazonaws.com/StagedProducts/Elevation/metadata/CA_CaliforniaGaps_B23/USGS_CA_CaliforniaGaps_B23_Project_Report.pdf).

## Actual measured accuracy versus specification

The same USGS project report provides **tested project-wide** results: DEM nonvegetated RMSEz 4.29 cm, corresponding reported 95% value 8.41 cm; vegetated 95th-percentile error 26.54 cm. Point-cloud nonvegetated RMSEz is 4.37 cm, reported 95% value 8.56 cm. These are stronger evidence than merely saying the dataset met a procurement specification. They are not measurements of every Newport street, nor narrow wall tops. No local checkpoint distribution or wall-specific test has yet been established.

The separate [18 April 2025 USGS acceptance summary](https://prd-tnm.s3.amazonaws.com/StagedProducts/Elevation/metadata/CA_CaliforniaGaps_B23/CA_CaliforniaGaps_2_B23/reports/USGS_CA_CaliforniaGaps_2_B23_Summary_Report.pdf) gives an expected open-terrain specification of 9.8 cm at 95% confidence, but directs readers to the project report for results. Its wording differs from the project report's required 10 cm RMSE/19.6 cm 95% table. Preserve that distinction; do not present the acceptance sentence as an independently measured local error. The report also limits classification verification to required classes.

The [NOAA metadata](https://www.fisheries.noaa.gov/inport/item/77693) reports a horizontal design class for Gaps 1/2 equivalent to approximately 0.509 m at 95%, rather than a verified 10 cm wall-position guarantee. With a narrow crest, horizontal displacement can select adjacent paving or a railing even where open-ground vertical accuracy is good. A 10 cm output raster would interpolate observations; it would not improve their underlying sampling or wall identification. The classified raw LAZ and along-wall cross-sections are the appropriate next evidence layer. Require continuous crest support, return classification and coverage checks; absence of returns is missing evidence.

## Useful older independent sources

The fresh catalogs also identify [2016 USGS West Coast El Niño lidar](https://www.fisheries.noaa.gov/inport/item/48222), whose inventory advertises nominal 0.12 m spacing. The targeted addendum found this conflicts with the original project description (approximately 0.35 m), so the smaller spacing must not be used as a dataset quality claim. Its coastal-strip footprints intersect the search envelope. A density advantage over 2023 is unproven, and the data predates the 2018 Main Island raising. See `2016-coverage-and-2005-validation-addendum-20261020.md` for actual small range-query checks: some Peninsula, Grand Canal and Little Island coverage exists; Main Island coverage is incomplete. Exact coverage at each wall must be checked from tile bounds/returns; a project-envelope intersection does not prove Island harbor-wall coverage. Use only to cross-check unchanged historical segments, with its original geoid/reference retained or transformed explicitly.

The inventory also lists 2012 NGA Los Angeles lidar, nominal 1 m spacing and advertised 3.5 cm vertical accuracy. The catalog explicitly disclaims USGS accuracy warranty. It is neither a demonstrated current-wall source nor proof of local survey accuracy; verify exact coverage and original report before considering it. The 2018 wildfire dataset contributes only one TNM tile in this envelope and is older than 2023. Neither should displace the current DEM based on a smaller advertised number alone.

## Tide reference: correct sign, still limited spatially

Fresh [NOAA metric datum metadata for station 9410580](https://api.tidesandcurrents.noaa.gov/mdapi/prod/webapi/stations/9410580/datums.json?units=metric) still reports accepted 17 April 2003, epoch 1983–2001, with MLLW 1.016 m and NAVD88 1.071 m above station datum. Thus a height expressed above MLLW converts to NAVD88 by subtracting **0.055 m**. The app's subtract-0.18-ft convention differs by only 0.000136 m from that published metric separation; rounding is not a meaningful explanation for the present simulation issues. At this station, 8.5 ft MLLW is approximately 8.32 ft NAVD88.

Crucially, [station details](https://api.tidesandcurrents.noaa.gov/mdapi/prod/webapi/stations/9410580/details.json) say the gauge was removed 27 January 1994. This historic station's accepted datum is useful, but it is not a present observed water-level feed. Its datum analysis period is 1980–1993, adjusted to the accepted epoch using Los Angeles control station 9410660. Do not use Newport predictions as measurements of present storm surge, pressure effects, or harbor gradients.

The [NOAA benchmark sheet](https://tidesandcurrents.noaa.gov/benchmarks/9410580.html) independently shows NAVD88 0.055 m above MLLW. It lists the historical highest observed water level on 28 January 1983 as 2.337 m above MLLW, about 7.67 ft. That gauge extreme is not a mapped flood extent or a current-wall overtopping threshold.

NOAA's [current epoch-update page](https://www.tidesandcurrents.noaa.gov/datum-updates/ntde/) still identifies 1983–2001 as operational and now proposes 2029 for the 2002–2020 products. Older NOAA factsheets referring to a 2025 release are outdated. Do not silently replace the model's epoch with 2002–2020 or add sea-level rise to an already datum-referenced observed tide without a defined scenario.

## Better spatial conversion, with uncertainty

[NOAA VDatum](https://vdatum.noaa.gov/download.php) currently offers software 4.8 and a 17 September 2025 regional package, including Southern California Inland version 01. It provides a reproducible route to evaluate MLLW–NAVD88 separation at harbor locations rather than assuming a single station offset everywhere. Before adoption, archive the grid/version and check coverage flags at actual boundary coordinates; compare transformed separation with 9410580's published benchmark tie. A tidal datum conversion does not calculate propagation delay or storm-water gradients.

NOAA's [uncertainty documentation](https://vdatum.noaa.gov/docs/est_uncertainties.html) reports Southern California Inland regional maximum cumulative uncertainty 6.7 cm as a **standard deviation**, not a local maximum-error bound. That figure combines an entire transformation chain and must not be applied automatically as the error of just MLLW–NAVD88 conversion. NOAA also warns that systematic subsidence/sea-level changes are excluded. The newer [West Coast spatial uncertainty report](https://vdatum.noaa.gov/download/publications/TM-61_VDatum_WestCoast_SVU.pdf) should be reviewed alongside the actual installed region; this audit did not successfully retrieve that PDF and does not assign its model to our harbor.

## Real-event evidence and what it can validate

[USGS documents the 10 January 2005 Newport Harbor flood](https://www.usgs.gov/media/images/january-2005-newport-harbor-flood): City photographs on Balboa Peninsula and Balboa Island were interpreted into observed inundation extents, and CoSMoS was compared against them. This establishes a genuine validation target rather than a visual plausibility check. The public illustration alone is insufficient as a calibrated polygon/time series. Retrieve original delineations, time-stamped photos and the Gallien et al. 2012 event forcing before computing skill scores. The 2005 walls also differ from post-2018 Island conditions, so reproducing the event requires historical barrier geometry.

The City's [April 2010 sea-level/flood-control report, printed p.57](https://www.newportbeachca.gov/home/showpublisheddocument/66959/637269446169670000) describes harbor flooding in January 1983 and 2005 with tides near 8 ft MLLW, and historical onset above 7 ft with tide gates closed. These are approximate, pre-raising descriptions. They cannot be used as a current Peninsula-wide crest or to force today's Island to flood at 7 ft.

## Actionable validation gates

1. Keep the 2023 original DEM as the regional terrain foundation; record its **tested project-level** error separately from sampling and hydraulic mesh spacing.
2. Acquire original project checkpoint locations/results and test independent Newport open-ground checkpoints before assigning a local terrain error. Do not use checkpoints used to fit the surface as independent validation.
3. Extract raw 2023 LAZ wall corridors and assess historical crest continuity against confirmed City/permit survey segments. Check 2016 coastal-strip coverage for independent unchanged-segment comparisons.
4. Build a versioned tide-reference audit using metric station ties and VDatum sample points. Resolve the undocumented January 2017 City survey vertical datum through original survey/control documents; agreement with one lidar point is not datum confirmation.
5. Validate a historical case only with matched event-era terrain/walls, observed boundary forcing, and time-referenced observed inundation. Test water-level/depth errors and wet/dry extent separately from arrival-time skill.
6. Preserve a present sealed-opening scenario as an explicit operating assumption. A no-flood result at ordinary tides is a useful falsification check, not sufficient validation of an 8.5 ft future/extreme scenario.